How AI can improve in video games
See patterns, make rules, reduce unnecessary thoughts.
AI in gaming is very widespread practice. You see it almost every video game that has come out since the inception of video games. In multiplayer games, you see these AI’s attempt to mimic human play. Take League of Legends as an example. There is a game mode, where you can replace the other players that would be in the Lobby with CPU players, and for someone learning the game this ends up being a debatably a good way to learn the game at the beginning. However, these bots eventually get outclassed very heavily, as a player gets better.
We see a similar thing with fighting games, and even games like chess in the earlier days of AI. Looking at chess, we have seen recently that chess now has AI’s such as Stockfish that depending on the amount of time it has to compute, can beat the best chess players in the world.
Today, I want to think about how AI can improve at games like League of Legends. There two types of situations that people already complain about in League of Legends related to CPU/AI. The first are derank or leveling bots that are programmed to do the bare minimum in terms of playing the game to not get caught. These are usually very bad at playing the game. The second is the idea of scripting, where people are aided by a script to perform actions either with inhuman reaction speed or inhuman accuracy. However, this still requires humans to make good decisions, though that is made easier with some scripts that provide enhanced knowledge such as the location of the enemy players on the map.
An AI that is really good League of Legends would not only be mechanically good at the game but also good at the game on a Macro level. And this is where I think it is good to look at how players view games like League and Chess. Often players create rules and patterns that they look for when playing. In chess, you have
Rule of the Openning
King safety
Control the center
Develope minor pieces
Knights on the rim are grimm
F2 and F7 are the weakest squares on the board
These are only a few examples. But many of the best players in the world have hundreds of rules and scenarios and patterns in their head that help them make decisions in game. They often will not have time to look at a board and calculate every move that can be made on the board, so they use their critical reasoning to ignore certain lines and in home in on others.
Similarly with a League of Legends player, they have ideas on how they can play the game with their champions. Junglers need to know where to be on the map at certain times, what side of the map to play one when they have certain compositions, what objectives are more important at different parts of the game. Laner similarly has rules on when to back, when to be aggressive, when to let certain wave states to happen, etc. What really separates great players from mediocre players are these rules.
With a game like chess it easier to do all the calculations required to figure out if a position is winning, but Stockfish employs these same principles. But a game like League of Legends has so many more variables, and so many unpredictable ones that this concepts of rules becomes even more important to eliminate, bad options, decisions and line quickly, and not even worry about them.
Just to add a Doctor Strange and Avengers reference, Doctor Strange did not need to look down every path to figure out which ones beat Thanos, he only needs to look at ones that would make the most sense, eliminate looking at paths that have events that would make victory close to impossible. When Doctor Strange, says there is one path that leads to victory, what he really means is that there are a set of rules and situations that they need to make sure they need to react in a certain way so they go toward the game states that are favorable and avoid the ones that are not. (Somewhat of Quantum mechanics way of looking at it. In terms of probabilities, rather a deterministic model)